Among the myriad protein interaction domains, PDZ domains are one of the most frequently encountered. They are often found in combination with other interaction modules and play an important role in establishing and maintaining cell polarity, in directing protein trafficking, and in coordinating synaptic signaling. Their importance is underscored by the severe neuronal and developmental phenotypes observed in PDZ domain knockout mice and by their implication in human congenital diseases. The enormous diversity of PDZ domain function is manifest in their abundance: there are at least 246 PDZ domains encoded in the mouse genome. To understand their individual roles, it is necessary first to define their recognition properties in a broad and biologically relevant fashion and to construct models that accurately predict their interactions across the mammalian proteome. Here, a combined experimental and computational approach is presented to detect, measure, predict, and perturb PDZ domain-mediated protein-protein interactions and to investigate their biological relevance. In Aim 1, a strategy is presented that uses protein microarrays and high-throughput fluorescence polarization to train a statistical model that predicts PDZ domain-peptide interactions on a proteome-wide scale. Experiments will initially focus on mouse PDZ domains, but will subsequently be extended to include human domains as well. Both the experimental data and predictive model arising from these efforts should prove valuable to the biological community that studies PDZ domain-containing proteins. In Aims 2 and 3, the biological relevance of novel interactions will be investigated in greater detail. Preliminary experiments have uncovered several novel interactions between PDZ domains and [unreadable]-catenin. In Aim 2, experiments are proposed to test the hypothesis that [unreadable]-catenin plays a functional role in forming or maintaining tight junctions. Preliminary experiments have also uncovered numerous novel interactions between PDZ domains and multidrug resistance proteins. In Aim 3, experiments are proposed to determine how PDZ domain-containing proteins modulate the cell-surface expression, subcellular localization, and function of these drug transporters. Understanding how multidrug resistance proteins are regulated may provide new ways to counteract common mechanisms of drug resistance. Finally, in Aim 4 a plan is presented to combine predictive modeling with protein microarray technology to design peptides and PDZ domains with tailored selectivities. As proof-of-concept, selective peptides will be designed to target PDZ domains that interact with [unreadable]-catenin or multidrug resistance proteins, and nonfunctional mutants of PDZK1 will be prepared to dissect its role in regulating blood cholesterol levels. Overall, it is our hope that both the novel interactions highlighted by our high-throughput experiments and the predictive models resulting from these data will be broadly adopted by the biological community as a powerful resource to study PDZ domain-mediated protein-protein interactions. PUBLIC HEALTH RELEVANCE: Almost all processes in biology are controlled by proteins that interact with each other in complex networks;errors in these networks underlie many developmental disorders, as well as human diseases such as cancer and autoimmunity. Here, we propose to use a high-throughput method developed in our lab to systematically study protein-protein interactions mediated by an important and prevalent family of protein modules known as 'PDZ domains', and then use the data resulting from these efforts to construct mathematical models that predict protein-protein interactions. This combined experimental and computational approach will provide the biomedical community with valuable tools to study PDZ domain-containing proteins and will teach us how to extend our efforts to include other families of interaction modules as well.